Chapter 10 - Line, circle, and ellipse detection

Abstract

This chapter starts by showing that lines may be located using inference methods such as the Hough transform (HT) and hypothesis-based methods such as RANSAC. The HT turns out to be especially valuable because of its robustness against noise, clutter, and partial occlusion. RANSAC uses pairs of edge points to generate hypotheses and selects the best one by checking the support provided by other edge points. The chapter also shows how the HT is able to locate circular and elliptic shapes and goes on to examine how all the parameters (five for an ellipse) can be estimated after location. A means for confirming that a shape is an ellipse is also presented. Finally, the chapter considers the problem of human iris location. Overall, the chapter emphasizes that the HT is highly robust because of its concentration on positive evidence for objects.